My PMM and I needed to launch a new feature for Razorpay's Wallet and Gift Cards app on Shopify. We needed a landing page—explaining the feature, who it's for, and how merchants could opt in.
He was crunched for bandwidth on Framer (marketing's tool). I had Replit access from work. I offered to build it.
We sat together and described what we wanted: hero with value prop, benefits breakdown, how it works, who it's for, user guide, FAQs, sign-up CTA.
Within minutes, we had a working page. It felt like magic.
Then came the design iteration.
The Iteration Reality
The first version was structurally sound. But visually? It needed refinement.
The pattern:
We couldn't give all changes upfront. Design iteration works by seeing something, reacting, adjusting. See, react, adjust, repeat.
But in LLM tools, each adjustment has a cost:
- Describe the change precisely
- Wait for generation
- Check if it matches what you imagined
- If not, describe again
After 10+ rounds: Time and tokens added up.
This is the hidden prompt tax. The cost in time and money of iterating on design through text descriptions instead of visual manipulation.
LLM Tool (Replit)
2-3 minutes per change
Visual Tool (Framer)
10 seconds per change
The Market Positioning Question
A few days later, I found Replit marketing "SaaS Landing Page Builder." Lovable with a page: "For Marketers."
LLM tools positioning landing pages as a core use case.
Here's what confused me:
Website building is the most mature no-code application. Marketers have been doing this for over a decade. Wix, Webflow, WordPress, Framer, Unbounce, Elementor. The options are endless.
Teams have learned these tools. Built workflows. Created brand libraries.
And these tools are adding AI features now. Framer has "Start with AI": describe page, AI generates, refine visually.
The question: Why would teams switch to LLM tools for landing pages?
The switching costs
| Friction | Impact |
|---|---|
| Learning curve for entire team | Weeks of ramp-up time |
| Multiple tool subscriptions | Cost doesn't permit redundancy |
| Prompt tax | Slower visual iteration |
| Loss of team autonomy | Only prompter can edit |
| No templates | Start from zero every time |
The use case doesn't fit LLM tools' strengths.
The Jobs to Be Done Mismatch
Marketing's JTBD for Landing Pages
- Fast visual iteration
- Team collaboration
- Template acceleration
- A/B testing built-in
- Brand consistency
What LLM Tools Offer
- Text-based iteration
- Generate from scratch
- Custom everything
- Flexibility for complex logic
- No templates needed
For standard landing pages, visual iteration is the job. LLM tools solve a different job: building custom logic from descriptions.
The Pattern That Matters
Building that Wallets page taught me a framework:
⚡ When a page needs to DO something
LLM tools fit:
- Calculate — ROI calculators, savings estimators
- Personalize — content adapts to user input
- Simulate — product configurators, interactive demos
- Integrate — custom API calls, real-time data
💬 When a page needs to SAY something
Visual tools fit:
- Explain — what the product is, how it works
- Persuade — why you should care, social proof
- Convert — clear value prop, strong CTA
My Wallets page was a SAY page: Explaining the feature. Persuading merchants why they need it. Converting them to install.
No calculations. No personalization. No custom logic. Just information architecture and visual design.
For SAY pages, visual tools with AI features (Framer's approach) make more sense. For DO pages, LLM tools shine.
The Smarter Hybrid Workflow
What works better:
Framer's "Start with AI" approach: AI for structure (60-70% there), GUI for refinement
Why LLM Tools Won't Just Add Visual Editors
Could LLM tools add visual editors? They might try. But visual iteration is Framer's core competency. Years perfecting components, real-time preview, collaboration features.
More likely: LLM tools niche down to where they're strongest—custom logic, complex software, DO pages. Visual tools keep improving visual iteration while adding AI for structure.
Both coexist. Different jobs to be done.
Where LLM Tools Actually Win
LLM tools aren't wrong for everything. They're excellent for DO pages:
| Use Case | Examples |
|---|---|
| Interactive product experiences | ROI calculators, product configurators, qualification checkers |
| Product simulations | Multi-step guided experiences, interactive demos with real calculations |
| Internal marketing tools | Sales intelligence systems, automated pitch generators |
| Unique campaign mechanics | Gamified interactions, custom integrations |
Templates can't handle these. Visual tools hit their limits. This is where LLM tools shine.
The Recommendation
For SAY pages (most landing pages)
Use visual tools with AI features:
- Framer's "Start with AI"
- Workflow: AI brief → generated structure → visual refinement
For DO pages (interactive, custom logic)
Use LLM tools:
- Replit, Cursor, Bolt
- When templates can't handle what you need
The Landing Page Brief Template
Whether you use Framer AI or LLM tools, this brief structure gets you 60-70% there:
| Section | What to Include |
|---|---|
| Target Audience | Who is this for? What's their pain? What do they need to understand? |
| Core Message | One sentence: What's the main point? Why should they care? |
| Page Structure | Hero (what/why), Benefits, How it works, Who it's for, FAQs, CTA |
| Brand Guidelines | Colors (with hex codes), Fonts, Tone |
| Key CTAs | Primary action, Secondary action |
| Constraints | Mobile-first, Load time, Accessibility |
| "Not Needed" (Critical) | What to explicitly exclude—no blog, no complex nav, etc. |
Example from my Razorpay Wallets page:
Want the one-page brief template?
The complete landing page brief structure—ready to use with Framer AI or any LLM tool. Built with the Razorpay PMM team.
Download the brief template